Messages in this thread | | | From | Vitaly Kuznetsov <> | Subject | Re: [PATCH RFC] memory-hotplug: add automatic onlining policy for the newly added memory | Date | Wed, 16 Dec 2015 11:53:35 +0100 |
| |
Xishi Qiu <qiuxishi@huawei.com> writes:
> On 2015/12/16 17:17, Vitaly Kuznetsov wrote: > >> Xishi Qiu <qiuxishi@huawei.com> writes: >> >>> On 2015/12/16 2:05, Vitaly Kuznetsov wrote: >>> >>>> Currently, all newly added memory blocks remain in 'offline' state unless >>>> someone onlines them, some linux distributions carry special udev rules >>>> like: >>>> >>>> SUBSYSTEM=="memory", ACTION=="add", ATTR{state}=="offline", ATTR{state}="online" >>>> >>>> to make this happen automatically. This is not a great solution for virtual >>>> machines where memory hotplug is being used to address high memory pressure >>>> situations as such onlining is slow and a userspace process doing this >>>> (udev) has a chance of being killed by the OOM killer as it will probably >>>> require to allocate some memory. >>>> >>>> Introduce default policy for the newly added memory blocks in >>>> /sys/devices/system/memory/hotplug_autoonline file with two possible >>>> values: "offline" (the default) which preserves the current behavior and >>>> "online" which causes all newly added memory blocks to go online as >>>> soon as they're added. >>>> >>>> Cc: Jonathan Corbet <corbet@lwn.net> >>>> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> >>>> Cc: Daniel Kiper <daniel.kiper@oracle.com> >>>> Cc: Dan Williams <dan.j.williams@intel.com> >>>> Cc: Tang Chen <tangchen@cn.fujitsu.com> >>>> Cc: David Vrabel <david.vrabel@citrix.com> >>>> Cc: David Rientjes <rientjes@google.com> >>>> Cc: Andrew Morton <akpm@linux-foundation.org> >>>> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> >>>> Cc: Gu Zheng <guz.fnst@cn.fujitsu.com> >>>> Cc: Xishi Qiu <qiuxishi@huawei.com> >>>> Cc: Mel Gorman <mgorman@techsingularity.net> >>>> Cc: "K. Y. Srinivasan" <kys@microsoft.com> >>>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com> >>>> --- >>>> - I was able to find previous attempts to fix the issue, e.g.: >>>> http://marc.info/?l=linux-kernel&m=137425951924598&w=2 >>>> http://marc.info/?l=linux-acpi&m=127186488905382 >>>> but I'm not completely sure why it didn't work out and the solution >>>> I suggest is not 'smart enough', thus 'RFC'. >>> >>> + CC: >>> yanxiaofeng@inspur.com >>> liuchangsheng@inspur.com >>> >>> Hi Vitaly, >>> >>> Why not use udev rule? I think it can online pages automatically. >>> >> >> Two main reasons: >> 1) I remember someone saying "You never need a mouse in order to add >> another mouse to the kernel" -- but we we need memory to add more >> memory. Udev has a chance of being killed by the OOM killer as >> performing an action will probably require to allocate some >> memory. Other than that udev actions are generally slow compared to what >> we can do in kernel. > > Hi Vitaly, > > So why we add memory when there is almost no free memory left? > I think the administrator should add memory when the free memory is low > or he should do something to stop free memory become worse.
I have virtual machines use-case in my mind where hypervisor adds new memory on high memory pressure reports from the guest (e.g. Hyper-V behaves like that). This is an automatic action.
> >> >> 2) I agree with Kay that '... unconditional hotplug loop through >> userspace is absolutely pointless' (https://lkml.org/lkml/2013/7/25/354). >> (... and I should had add him to CC, adding now). Udev maintainers >> refused to add a rule for unconditional memory onlining to udev and now >> linux distros have to carry such custom rules. >> > > If the administrator don't know how to config the udev, he could use sysfs > (echo 1 > /sys/devices/system/node/nodeXX/memoryXX/online) to online it, > or write a script to do this.
Oh, no, I'm not taking about manual actions here. My suggestion doesn't eliminate this possibility and it doesn't even change the default -- memory blocks stay in 'offline' state unless someone requests the auto-online policy.
-- Vitaly
| |